Vehicle Type Heterogeneity in Battery Electric Vehicle Energy Consumption: A Multi-Perspective Analysis Using Real-World Fleet Data

Real-world energy consumption of battery electric vehicles (BEVs) varies with trip conditions, driving behavior, and vehicle state. Treating vehicle type merely as a control variable may mask substantial heterogeneity in energy consumption mechanisms. This study treats vehicle type as a source of mechanism heterogeneity and examines how the determinants of energy consumption rate (ECR) differ among sedans, logistics vehicles, and buses. Using 41,933 trips in China, we combined statistical analysis, machine learning, and CausalForestDML. Statistically, ECRs of sedans and logistics vehicles decrease as trip distance increases, but trip distance is nearly irrelevant for buses. Aggressive driving raises logistics vehicle ECR by 22.2% yet barely affects sedans. For prediction, permutation importance further reveals the divergence: trip distance and duration dominate sedan prediction, whereas energy recovery ratio and accelerator-pedal, which are marginally important for sedans, become primary predictors for logistics vehicles and buses. Causal estimates confirm that directional divergences persist after conditioning on covariates. For example, trip duration has opposite effects on buses versus sedans and logistics vehicles. Moreover, the causal framework decomposes each factor’s effects across its value range, revealing patterns beyond correlation alone. These findings provide actionable guidance for vehicle-specific eco-driving strategies and battery management system optimization.